Eddy covariance measurements and parameterisation of traffic related particle emissions in an urban environment

HAL Id: hal-00301677

https://hal.archives-ouvertes.fr/hal-00301677

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Eddy covariance measurements and parameterisation of

traffic related particle emissions in an urban

environment

E. M. Mårtensson, E. D. Nilsson, G. Buzorius, C. Johansson

To cite this version:

E. M. Mårtensson, E. D. Nilsson, G. Buzorius, C. Johansson. Eddy covariance measurements and parameterisation of traffic related particle emissions in an urban environment. Atmospheric Chemistry and Physics Discussions, European Geosciences Union, 2005, 5 (4), pp.5541-5584. �hal-00301677�

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

Atmos. Chem. Phys. Discuss., 5, 5541–5584, 2005 www.atmos-chem-phys.org/acpd/5/5541/

SRef-ID: 1680-7375/acpd/2005-5-5541 European Geosciences Union

Atmospheric Chemistry and Physics Discussions

Eddy covariance measurements and

parameterisation of tra

ffic related particle

emissions in an urban environment

E. M. M ˚artensson1,2, E. D. Nilsson2, G. Buzorius3, and C. Johansson2,4 1

Department of Meteorology, Stockholm University, S-106 91 Stockholm, Sweden 2

Section for Atmospheric Science, Department of Applied Environmental Research, Stockholm University, S-106 91 Stockholm, Sweden

3

Department of Research, CIRPAS, Naval Postgraduate School, CA, USA 4

Environment and Health Protection Administration, Box 38 024, S-106 64 Stockholm, Sweden

Received: 9 May 2005 – Accepted: 24 June 2005 – Published: 29 July 2005 Correspondence to: E. M. M ˚artensson ([email protected])

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

Abstract

Urban aerosol sources are important due to the health effects of particles and their potential impact on climate. Our aim has been to quantify and parameterise the ur-ban aerosol source number flux F (particles m−2 s−1), in order to help improve how this source is represented in air quality and climate models. We applied an aerosol

5

eddy covariance flux system 118.0 m above the city of Stockholm. This allowed us to measure the aerosol number flux for particles with diameters >11 nm. Upward source fluxes dominated completely over deposition fluxes in the collected dataset. There-fore, the measured fluxes were regarded as a good approximation of the aerosol sur-face sources. Upward fluxes were parameterised using a traffic activity (TA) database,

10

which is based on traffic intensity measurement.

The footprint (area on the surface from which sources and sinks affect flux measure-ments, located at one point in space) of the eddy system covered road and building construction areas, forests and residential areas, as well as roads with high traffic den-sity and smaller streets. We found pronounced diurnal cycles in the particle flux data,

15

which were well correlated with the diurnal cycles in traffic activities, strongly support-ing the conclusion that the major part of the aerosol fluxes was due to traffic emissions. The emission factor for the fleet mix in the measurement area EF_{f m}=1.4±0.1×1014 veh−1km−1was deduced. This agrees fairly well with other studies, although this study has an advantage of representing the actual effective emission from a mixed vehicle

20

fleet. Emission from other sources, not traffic related, account for a F₀=14±18×106m−2 s−1. The urban aerosol source flux can then be written as F=EF_{f m}TA+F₀. In a second attempt to find a parameterisation, the friction velocity U_∗ normalised with the aver-age friction velocity U_∗ has been included, F=EF_{f m}T A

 U_∗ U_∗

0.4

+ F0. This

parameter-isation results in a somewhat reduced emission factor, 1.3×1014veh−1 km−1. When

25

multiple linear regression have been used, two emission factors are found, one for light duty vehicles EF_LDV=0.3±0.3×1014veh−1km−1 and one for heavy-duty vehicles,

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU EF_HDV=19.8±4.0×1014veh−1 km−1, and F₀=18±16×106m−2 s−1. The results show

that during weekdays ∼70–80% of the emissions came from HDV.

1. Introduction

Anthropogenic aerosol particles have the potential to influence climate directly through Mie-scattering and absorption and indirectly when they serve as cloud condensation

5

nuclei (CCN). Of the anthropogenic climate forcing components, the impact of the aerosol is far more uncertain in magnitude than the greenhouse gases, but potentially of equal magnitude (IPCC, 2001). Recently, the aerosol loading expressed in number concentration of fine particles has been related to adverse health effects in epidemi-ological studies (WHO, 2003, 2004; Ibald-Mulli et al., 2004). It is therefore important

10

that we in an accurate and realistic way can represent aerosol sources in both local to regional air quality models and regional to global climate models. Measurements of the aerosol emission fluxes can contribute in two important ways: They can allow us to develop accurate and efficient parameterisations of the sources, and they can help validate the emission inventories used in models. Such knowledge is necessary for

15

assessing abatement strategies that aim at reducing particle emissions in cities. In the urban environment, one of the main aerosol sources is the road traffic (Ru-uskanen et al., 2001, and Gidhagen et al., 2005), which produces aerosols both due to the combustion of fossil fuels and due to mechanical processes, e.g. wear of break linings, tires, and road surface. To be able to include the traffic aerosol emissions in

20

a model, we must at least be able to relate the emitted aerosols to traffic intensity. In addition, one should try to relate meteorological conditions, vehicle speed, and type of engine or fuel to the aerosol emissions quantitatively, in order to derive a source param-eterisation, or different emissions factors for different conditions. This may be achieved through measurements of individual cars in the laboratory (e.g. Maricq, 1999), road

25

tunnel studies (Gidhagen et al., 2003; Kristensson et al., 2004), by tracking individual cars in the traffic (Kittelsson et al., 2004), or by estimates based on roadside

measure-ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

ments (Gidhagen et al., 2004a, b; Ketzel et al., 2003). Street canyon and tunnel studies provided a significant development in urban pollution understanding by presenting pol-lution rates from assembles of numerous cars operating in situ. Such methods have considerable limitations and uncertainties when used to represent the actual effective emissions from a whole mixed vehicle population in an urban area.

5

The eddy covariance (EC) technique allows us to obtain an integral measurement of the amount of particles emitted from a wide area in the city (the footprint area could be as large as 1–10∼km2) covering all sources and sinks in their natural environment. Such measurements are still scarce (Dorsey et al., 2002) but nevertheless, very useful since they give a direct measurement of the pollution source strength.

10

The aim of the present study is to use the eddy covariance method and thereby direct in situ flux measurements representative of the true traffic mixture covering a large geographic area to derive emission factors and if possibly a source parameterisation.

2. Measurements

2.1. Campaign logistics and site description

15

During 49 days, from 19 March to 6 May 2002, the turbulent aerosol particle number fluxes were measured in Stockholm, the capital of Sweden and the largest city in Swe-den. The city has 750 000 inhabitants and the county of Stockholm has 1.8 million inhabitants. The measurements were made from the top of a telecommunication tower in the southern central part of the city. The tower is built in concrete, hexagonal in

20

shape and 105 m tall and located 28 m above the sea level. On the top of the tower is a 10-m tall framework. The measurement equipment was placed at the top platform of the framework, and the sampling was ejected above the platform in order to separate it from the concrete construction and minimize the flow disturbance from the tower.

The surrounding landscape varies with the point of the compass see Fig. 1. For

25

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

with the size 1 km×1 km and the type of land cover will be described. This is partly motivated by the grid resolution of the traffic database, which also contain information on surface properties. For a more detailed analysis later on, each of these sectors are then divided into two sectors with the width of 45◦, totally eight sectors.

The North East sector (NE) covers 0◦–90◦. Fifty meter north of the tower a road with

5

high density of traffic (32 000 vehicles per day) passes in easterly direction. Beyond the road, (100–400 m) the area consists of blocks with offices, shops, streets, and car parks. Approximately 400 m from the tower is a road and building construction area. The area between 700 to 1000 m from the tower is an industrial area. Between the construction area and the industrial area there are water and a small patch of a forest.

10

The heights of the buildings are 5–15 m.

The South East sector (SE) covers 90◦–180◦. The area closest to the tower consists of residential areas and smaller streets. Further away is a forest. The heights of the buildings are 15 m and the forest canopy (mainly spruce and pine) is 10–15 m high.

The South West sector (SW) covers 180◦–270◦. Between 180◦ and 210◦ is a

resi-15

dential area. From 210◦ up to 270◦ an approach (ca. 85 000 vehicles per day) passes through the sector in north-north-westerly direction at a distance of 800–900 m from the tower. From 225◦ to 270◦ another road with dense traffic pass between the tower and the highway. Further away in the south-south-westerly part roads with dense traffic pass through the sector. Between the tower and the highway are residential areas. In

20

the 220◦–250◦directions from the tower and beyond the highway are large sport arenas located. In the same area are also stores, offices, and a shopping centre located. Be-yond the highway, at approximately 265◦–270◦is a roundabout with high traffic intensity. The heights of the buildings in the S.W. sector are 10–20 m.

The North West sector (NW) covers 270◦–360◦. The approach and the larger road

25

from the SW passes over bridges (62 000 vehicles per day) in the direction of 270◦– 300◦ and run into a tunnel. The same road as in the NE passes close to the tower in a westerly direction. Beyond this road are blocks with offices, shops, streets, and car parks, similar to those in the NE sector. Construction of a new tramline was performed

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

in the area during the measurement campaign. At the other side of the water area is an island. This part is narrower to the city centre and has more of a city centre structure, with residential areas, restaurants and shops. Many offices are located in this sector. Some of the streets have high traffic intensity. The heights of the buildings are 10–30 m. 2.2. Eddy covariance method

5

The vertical aerosol number flux was calculated with the eddy covariance technique. If we separate the aerosol number concentration N and the wind speed w in mean and turbulent fluctuations;

N= N + N0 and w = w + w0 (1)

where the overline denotes temporal mean and primes 0 the turbulent deviations

10

from the mean, and use the Reynolds decomposition we can write the vertical flux;

wN = wN + w0N0 (2)

where wN is the mean wind transport by low frequency large-scale transport and w0N0

is the turbulent eddy transport. The turbulent flux is the covariance between fluctua-tions in vertical wind and in concentration, and it is the net transport that results from

15

both sinks and sources at the surface, under the assumption that the measurements are made within the surface layer or “constant flux layer. Sensible heat w0_T0 _and

mo-mentum fluxes w0u0+ w0v0 are calculated in a similar manner. The fluxes were calcu-lated over periods of 30 min, to include the largest eddies with slow fluctuations and still exclude long-term trends in the data. The fluctuations w0 and N0 were separated

20

from the mean by linear detrending, which also removes the influence of low frequency trends. Prior to the calculation of w0N0 the wind vectors where rotated to achieve a zero w0, and the aerosol data was shifted in relation to the wind data to correct for the time lag in the sampling line (calculated from the maximum correlation).

The three dimensional wind and temperature were measured at 20 Hz using a Gill

25

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

measurement platform, in total located 118.0 m above the ground. The aerosol was sampled just beneath the sonic head through a 3.83 m sampling line of 1/4-inch stain-less steel.

To measure the total aerosol number concentration (D_p>11 nm) we used a

Conden-sation Particle Counter (CPC), model TSI 3762, which was logged through a

pulse-5

to-analogue voltage converter and one of the Gill R3 external analogue signal input lines. The sampling flow Q, through the CPC was 3.083 lmin−1 but a bypass flow of 4.0 lmin−1 assured turbulent flow in the sampling line. For maximum accuracy, the correction for coincidence was determined by using the equation from TSI (2002),

N_a=N_i exp (N_aQτ_p), where N_a is the actual number concentration, N_i is the indicated

10

number concentration (particles/cm3), Q=51.4 cm3 s−1 and τ_p=0.2 10−6s is the ef-fective times each particle resides in the viewing volume. N_a in the exponent can be approximated by N_i.

The frequency response of the CPC is limited, its first-order response time constant

τ_ccan be tested (Buzorius, 2001), for our CPC model is τ_c∼0.4 s. The underestimation

15

of the flux due to limited sensor response depends on the frequency f of the turbulence, which is determined by the observation level z, mean horizontal wind speed U and stratification z/L, where L is the Obukhov length. If the slow response time is the principal reason for underestimating the flux, as in our case, this underestimation can easily be calculated (Buzorius et al., 2000) according to

20 F_m F = 1 1+2πn_mτ_cU_z α, (3)

with α=1 for stable stratification and α=7/8 for neutral and unstable stratification. In Eq. (3) n_m is the normalized frequency

n_m = 2.0 − 1.915

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

for stable stratification, with the neutral limit n_m=0.085 applying also to the unstable case.

The aerosol fluxes presented in this work has been corrected according to Eqs. (3) and (4). The mean correction caused by the frequency response was 8% and the median correction was 3%. Owing to the stochastic nature of the turbulence (Rannik

5

and Vesala, 1999) and the discrete counting of aerosols (Buzorius et al., 2003) there is always an uncertainty δ(w0N0) in the time averaged flux estimates. The stochastic nature of turbulence causes an uncertainty of about 10%. According to Buzorius et al. (2003), the uncertainty in the flux, due to the discrete counting can be expressed as

δw0_N0 = σwN

q

NQ∆t

(5)

10

where σ_w is the standard deviation of the vertical wind, N is the aerosol number con-centration averaged over the sampling period ∆t (in our case 30 min), and Q is the sampling volume flow rate through the particle counter. The discrete counting error averaged 0.54%, the median 0.38% with the 25–75 percentiles at 0.17 and 0.62%, in summary very small errors owing to the high aerosol number concentrations. The

15

errors due to tube damping are also negligible owing to the relatively short tube length and since we have ensured the flow to be turbulent by adding the extra bi-pass-flow. 2.3. Traffic activity

A grid resolved database is available with the hourly traffic activity expressed as vehicle-kilometres per time unit. This database is part of a regional air quality

man-20

agement system operated by the local environmental authority in the city of Stockholm and includes all types of air pollutant sources in the counties of Stockholm and Up-psala (30 different municipalities). Traffic data are based on traffic measurements on different road types during different years. The whole database is updated once a year.

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

We have used the database for 2001 to obtain the traffic activity during the flux mea-surements March-May 2002. No direct real-time traffic counters were available for the measurement period.

However, from long-term observation one can expect that year-to-year variations are negligible except in areas with major changes in traffic pattern or construction areas.

5

From 2001 to spring time in 2002, no changes in traffic pattern were reported within the spatial scales described in Fig. 1, but new constructions emerged within the footprint area in one of the sectors as it is described in this paper. While we lack point measure-ment of the traffic activity during the exact campaign period, we have instead a great advantage in a traffic database that gives an information representative for the average

10

amount of traffic (vehicle kilometres) for all roads within the whole measurement sector. The location of the measurement site in this study gives the possibility to measure fluxes from areas with different land cover, e.g. areas with densely trafficked highways, one area mainly covered with forest and residential areas. The types and number of vehicles in these areas differ over time (weekdays and holidays). The relative amounts

15

of heavy-duty vehicles (HDV) and light duty vehicles (LDV) is potentially very important for the emissions as HDV emit more particles since they are diesel fuelled (the traffic in Stockholm has ∼6% HDV). The vehicle speed is another factor that influence the emission. At higher speed, the vehicles emit more particles (Kittelson et al., 2004). 2.4. Complementary measurements

20

Total particle number concentrations have been measured at a roadside location, Hornsgatan (street canyon), and at an urban background location, Rosenlundsgatan (Gidhagen et al., 2003). Hornsgatan is a 24 m wide, four-lane street surrounded by 24 m high buildings on both sides, thus being a rather symmetric street canyon with a unity width/height ratio. Traffic intensity is about 35 500 vehicles per day during

week-25

days, with an average of 5% of heavy-duty vehicles, mostly buses of which almost all uses ethanol for fuel. Of the light duty vehicles, there is an average of 5% diesel fu-elled cars, mainly taxis. Air intakes are placed on a trailer 1.5 from the fac¸ades, at

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

3 m height above the street surface. A CPC3022 instrument (TSI Inc.) was used to measure total number concentrations (D_p>7 nm). An identical instrument is located at

the urban background station, Rosenlundsgatan, about 600 m from Hornsgatan and at 30 m height on a roof.

3. Results and Discussion

5

3.1. Average values and covariances

During the campaign positive upward aerosol fluxes were most frequent (88% of the time) and roughly an order of magnitude stronger (average value 299×106m−2 s−1) compared to downward negative (deposition) fluxes (average value 26×106m−2 s−1). Net deposition mainly took place during night time and times when the traffic was low.

10

In the parkland of the SE sector net deposition fluxes occasionally occurred also during daytime. The selection of the data set representing net deposition was studied in detail within the project in a Master thesis (Brokh ¨oj, 2003), and evaluated in conjunction with aerosol deposition models, Schack et al. (1985), Slinn (1982), and Zhang et al. (2000). The net upward fluxes were naturally also affected by deposition, although the upward

15

source fluxes dominated. However, the fact that the net upward fluxes are an order of magnitude larger than the net downward fluxes indicates that the deposition can be neglected when studying the net upward fluxes. The comparisons with the three well-established aerosol deposition models further support this conclusion. In conclusion, the error we make by neglecting the deposition when we consider the net upward fluxes

20

as true source fluxes is approximately 10%.

Table 1 (Table 1a for all sectors and Table 1b for the SW and SE sectors) summarizes the mean characteristics of key parameters measured by the EC system during the campaign, and the correlation between these parameters. Note, that the aerosol flux is better correlated with friction velocity U_∗ (R=0.449) than with the average aerosol

25

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

height the turbulence at the site is an important factor for the vertical flux exchange and confirms that deposition fluxes are probably less important (since they should be dependent on aerosol number). The average number concentration is instead strongly influenced by source regions located outside the footprint A small anti correlation for the aerosol with temperature can be seen (R=−0.171). Previous studies have seen

5

higher aerosol concentrations at street level at low temperatures and attributed them to higher nucleation rates in the exhaust (Gidhagen et al., 2003). The traffic activity has the highest correlation with the aerosol flux w0_N0 _{(R=0.788), when the southerly}

sectors are used, higher than for any of the meteorological parameters in Table 1b. More discussion about this will follow later.

10

Figure 2a–h demonstrates an eight-day long period of temperature, sensible heat flux, horizontal wind speed, friction velocity (closely related to momentum flux), stability (z/L, where L is the Obukhov length and z is the measurement level), aerosol particle number flux, and aerosol number concentration and wind direction. Temperature data showed a clear diurnal cycle on most of the days being 5 to 10 degrees higher during

15

the day compared to the night. Daily averages of air temperature increased during the campaign from the 265 to 280 K in the end of March up to 277 to 287 K in the end of April.

Turbulent heat fluxes exhibited strong diurnal variation on most of the days. Horizon-tal mean wind speed was varying from 0 to 10 m s−1(12 m s−1if considering the entire

20

campaign). Friction velocity being a derivative of the momentum flux varies from the 0 up to about 1 m s−1 with a diurnal cycle similar to that of heat flux. Particle fluxes exhibited strong diurnal variation on most of the days. Presumably, heat flux diurnal variation is caused mainly by the diurnal cycles in solar radiation flux and atmospheric turbulence whereas particle flux depends primarily on traffic intensity within the

foot-25

print area, which varies with wind direction. For instance, on sunny weekend days heat flux has similar values to those on sunny weekdays, while particle flux is tremen-dously different on these days (weekend example days 110 and 111 in Fig. 2, compare Figs. 2b and f. Figure 2g shows that the aerosol number concentration ranging from

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

the 103to 2.5×104particles cm−3for the tower.

Concentration was measured in three places: above the city at 118 m altitude from the surface, inside the street canyon at 3 m level from the street surface and at rooftop level near the street level measurement site (Fig. 3). The roof and street level mea-surements were located in the northwest direction about 2 km from the tower. During

5

some periods the correlation in aerosol time series and similarity in absolute values at the rooftop and tower sites are remarkable despite the spatial distance. Buzorius et al. (1999) demonstrated similarly high spatial homogeneity in aerosol number con-centration over the distance of 2 km in Helsinki (R>0.8) where traffic is the dominant aerosol source. Concentration time series exhibited maximal values during the day

10

and minimum values during the night. The highest aerosol concentrations were mea-sured in the street level, near the actual traffic source. Aerosol number was decreasing as air from the street was transported to roof level and tower level. Presumably, the mechanisms decreasing the aerosol number were mainly ventilation/dilution and to a small extent dry deposition and coagulation. Calculations presented by Gidhagen et

15

al. (2004a) have shown that on average the effect of coagulation and dry deposition is expected to be small (<10%) at the levels encountered in this urban area. The ra-tio between the roof and tower level aerosol number concentrara-tion is in most of the cases larger than unity showing that the tower site is measuring less particles than at the rooftop site, which is closer to the aerosol sources-traffic. The ratio reaches larger

20

values (varying from 0.5 to 12) at the night and shows less variability during the day (from 0.7 to 3.5) due to the unstable atmosphere. During the stable stratification (i.e. night-time), the atmosphere is less mixed and larger variability in aerosol concentra-tion is expected compared to the mixed atmosphere. Addiconcentra-tionally the difference in time series result from different aerosol sources located in different footprint areas. If wind

25

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

3.2. Spectra and co-spectra

Figure 4 demonstrates samples of 20 Hz raw data. The data was acquired on 5 April 2002, from 13:00 to 13:30 local time. The half an hour time period was chosen ar-bitrarily. Averaged horizontal wind speed was 8.3 m s−1, standard deviation 1.78 m s−2 and friction velocity during this time period 0.87 m s−1. The atmosphere was

un-5

stable with the Obukhov length being −353 m. On average, the particle counter was measuring 7×103particles cm−3 with a standard deviation of 2×103. Averaged wind direction during the half an hour time period was from North (20◦). Eddy covariance calculations resulted in 1200×106particles m−2s−1and 170 Wm−2for aerosol and heat fluxes, respectively. Particle flux was larger during this episode compared to the study

10

average values but has a typical value for this wind direction. Figure 4a demonstrates a high variability in aerosol concentration time series. Number concentration changes within seconds more than 100%. Aerosol particles emitted from the individual sources travelling towards the sampling point have not mixed well with air parcels, which have trajectories that have not crossed the sources. It is interesting to note how at 1000 to

15

1100 s temperature and aerosol concentration was reduced and simultaneously verti-cal wind speed (Fig. 4c) was reduced from 2–4 m s−1to close to zero. Before the drop, air was coming from below the measurement altitude bringing the warmer and more polluted air, which was replaced by the relative cleaner air at around 1060 s.

All the raw data for each half hour period during the entire campaign was

Fast-20

Fourier-Transformed for spectral and co-spectral analysis. Results showed typical di-urnal variability in the spectra – containing more energy at lower frequency range dur-ing the day compared to the nigh time. In general, spectra and co-spectra obeyed the −2/3 and −4/3 power decay laws during unstable atmospheric stratification. However, aerosol power spectra exhibited slightly faster decay than the −2/3. This is due to the

25

limited CPC response time, attenuation in the sampling lines etc. The effect would have been greater at a lower height, but the fluxes is in any case corrected for the attenuation by the CPC using Eqs. (3) and (4). Figure 5 presents samples of spectra

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

calculated using the raw data shown in Fig. 4.

At the low frequency range, all spectra decayed with peaks around 0.1 to 0.3 s−1 pe-riodicity. The example of spectra in Fig. 5 (when energy decay with the lower frequency left of the peak) is considered to be stationary. There were many occasions when a clear peak could not be found by visual inspections of the spectra. Instead, the largest

5

energy was contained in the lowest frequency range. Dorsey et al. (2002) reported a similar mixture of stationary and non stationary-distorted spectra in the data sampled in an urban environment.

Figure 6 demonstrates co-spectra of vertical wind speed and (a) temperature, (b) aerosol number concentration. The integral under the co-spectra equals the total flux

10

value. Most of the co-spectra values are positive indicating positive correlation and hence upward fluxes. Both co-spectra are in the inertial range decaying according to the −4/3 slope, which agrees with scaling laws. Although the aerosol power spectra in Fig. 5c decayed faster than the −2/3 indicating attenuation of the signal. Figure 6 show that the loss in the signal was not correlated with the vertical wind speed and hence

15

had no influence in the co-spectra. Aerosol co-spectra decays with −4/3 as theory predict. Analysis on spectra and co-spectra concluded that atmospheric turbulence at the sampling site obeyed scaling laws and eddy covariance measurement represent vertical exchange fluxes.

3.3. Sources and wind direction

20

The footprint area influences the magnitude of the aerosol particle flux. As shown in Fig. 1 the areas around the tower consist of different types of land cover. Fluxes from all directions around the tower are to various degree included in the data set. Variations in the wind direction is therefore one of the reasons for the variability in the aerosol flux. The fluxes are hence sorted in eight groups depending on wind direction. Each sector

25

is 45◦ broad. The weekdays (Monday to Friday) and holidays (Saturday, Sunday and public holidays) are treated separately. For the fluxes the mean, the median, and 25 and 75 percentiles are calculated, see Fig. 7a–b.

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

Figure 7a shows the averaged fluxes for the 32 weekdays. The smallest fluxes with less variation were found in the sectors between 90◦and 225◦. These are the sectors with residential areas, forest and only minor streets. The NE sectors had the largest mean fluxes (several hundred millions particles per square meter and second) with the highest variation. In these sectors, the footprints must have contained both the

5

construction area and the roads with dense traffic. From the sectors between 225◦ and 360◦, the fluxes were also high with a large variation. Through these sectors, an approach with dense traffic passes and the construction work was going on between 315◦and 360◦.

Weekends and holidays have different particle flux characteristics. In Fig. 7b, the

av-10

erage fluxes during the 17 holidays are shown. The lowest fluxes came from the same sectors as during weekdays with the forest and residential area. The largest fluxes came from the areas that included larger roads. From the areas with construction work that had the largest fluxes during weekdays, the fluxes were reduced. The variations in the fluxes were much smaller during the holidays. The reason for this can be a uniform

15

mixture of vehicles, presumably with less HDV, and the variation in the speed is less because of the absence of rush hour periods with queues of cars. In the construction area the activity is smaller than during the weekdays. It appears from this first com-parison that the magnitude of the aerosol source fluxes are consistent with the type of sources present in different directions assuming that traffic is the strongest available

20

source.

3.4. Traffic and wind direction

The traffic database includes information about all traffic activity on the roads in the area, from the small streets in the residential area to the large highway. This spa-tial distribution together with the diurnal and weekly variation gives an opportunity to

25

account for all traffic related sources influencing the emissions.

The hourly average traffic activity per area and time unit TA (veh km m−2 s−1) is estimated for the same eight 45◦broad sectors as the fluxes, Fig. 1. The traffic activities

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

are available for both LDV and HDV, separately. Figure 8 shows the averaged traffic intensity (bars) for the eight sectors during the spring of 2001. The traffic is expected to be the same in 2002 since no significant changes had occurred in the traffic system. The gray bars show the LDV and the black bars the HDV. During holidays the total

TA is reduced, the LDV traffic during holidays is 77% of the weekday TA and the HDV 5

traffic during holidays is 35% of the weekday HDV. The fraction of HDV of the total

TA differs between time and sector from 2% during holidays to 8% during weekdays.

The similarities between Figs. 7 and 8 supports the conclusion that the differences in flux for different wind directions are largely due to differences in traffic related sources. However, the agreement is not perfect indicating that other factors also matter. Please

10

note that the traffic database does not consider the changed activities in the sector with the construction work, where many extra transports in and out of the area were carried out. Within this area a lot of extra of road vehicle work was performed.

3.5. Diurnal cycles

The aerosol flux from the four sectors are diurnally averaged over the campaign for

15

either the weekdays or the holidays, see Fig. 9a–d, the times are in Local Swedish Time. During weekdays, the fluxes increased sharply between 05:00 and 08:00 in all sectors with some differences between the sectors. In the SW sector the increase came earliest and in the SE sector latest, the increase is strongest in the NE sectors. The emissions during daytime have different patterns in the four sectors. In the SW

20

sector the maximum occurs 08:30, in the NW sector around 09:30 and in the NE sector around 10:30, reaching values of about 5−12×109m−2 s−1. In the cleaner SE sector, there are two modes one at 12:00 and one at 15:00 with fluxes <2×108m−2s−1. In this sector, the variation over the day is less than in the other sectors. The fluxes decrease at 17:00 in the NE and SW sector, but in the NW sector not before 20:00. For all sectors

25

except SE, the fluxes have a large variation during the day. The reason for this can be differences in the diurnal traffic activity. The high fluxes in the morning coincide with the time of the rush hour. Fluxes during the afternoon rush hour are less pronounced. In

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

the middle of the day, however the explanation for the relatively high emissions can be more transports carried out with HDV, due to deliveries within the areas and transports passing through, and from off-road diesel fuelled machinery. Another factor can be that with fewer cars, the driving speeds increase, and together with the increased HDV, this may result in higher emissions. Of possible importance are also the more turbulent

5

conditions at noon when the boundary layer is fully developed. The results are in agreement with the measurements made by Dorsey et al. (2002), who observed a mid day maximum in aerosol fluxes in Edinburgh.

During holidays the diurnal cycles showed a different behaviour for three of the sec-tors, while in the SE sector the cycle is similar as during weekdays. The emissions

10

were considerably lower for all sectors, the increase in the flux values during the morn-ing was less pronounced and the maximum occurs later on the day. This is expected, as the heavy traffic is considerably lower and most people do not have to drive to their work places on holidays. The traffic is motivated more by shopping and pleasure

The average diurnal cycles during weekdays for the particle number concentration,

15

friction velocity and heat flux during weekdays are included in Figs. 10a–b. Only the two southerly sectors are shown. The average background concentrations are between 2500 and 3500 cm−3 around 03:00. In the morning at 08:00 a maximum in number concentrations for the day occurs. We can see a correlation with the fluxes, but the average number concentration N is more effected by the stable stratification in the

20

morning, lower height of the boundary layer and less efficient dilution, which results in an N maximum. Friction velocity increases at the same time as the fluxes and have a maximum between 12:00 and 15:00 depending on the sector. The heat flux is negative before 07:00 and after 19:00 and has a maximum in all sectors around 13:00.

Figures 11a–b show the traffic activity (LDV and HDV) compared with the particle

25

fluxes in the SW sector for weekdays and holidays. On weekdays, Fig. 11a, the LDV traffic has two peaks, one at 08:00 and another at 17:00, this is similar for all sectors. The HDV traffic has more the shape of a mode centred at 11:00. The increase in traffic (HDV and LDV) coincides with the rise in fluxes and in the afternoon, the fluxes

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

decrease at the same time as the reduction in the traffic activity occurs. The peaks in LDV coincide with high fluxes and at noon, the fluxes seem to be more influenced by the HDV. In the evening, there are sometimes increased fluxes at the same time as the HDV increases. For holidays, Fig. 11b, the LDV has one maximum at noon and the HDV has one morning and one afternoon peak. In the morning the fluxes

5

coincide with the increased traffic, the afternoons cannot be compared due few data points. At holiday nights and at noon the LDV is higher compared to the weekdays. For all sectors, except the NW higher emissions were observed during holiday nights than during weekday nights. There was sometimes even a small maximum in the evening. The higher fluxes during the holiday nights as compared to weekday nights can bee an

10

effect of the higher LDV especially the usage of taxi cars during the weekends, since these cars are often diesel fuelled.

It appears that the averaged diurnal cycles of the aerosol fluxes also support the suggestion that most of the aerosol source is related to traffic, even if we can see the influence of the turbulence when the fluxes are compared with the friction velocity.

15

Especially in the middle of the day, when the fluxes show a strong correlation to the friction velocity, this dependency might have to be considered.

3.6. Source parameterisation

According to previous discussion regarding the deposition fluxes, these can be ne-glected and the aerosol number source fluxes can be approximated with the measured

20

positive upward fluxes w0_N0_{. In Fig. 12 the hourly averaged source fluxes (particles}

m−2s−1) from wind directions between 90◦and 270◦ (four 45◦wide sectors, Fig. 1) are shown as functions of TA. In total 350 data points are included from low to high TA. The north sectors are excluded due to the construction works in these sectors. Our intention is to find a first order parameterization for the urban source flux F . A linear fit

25

to the data gives (see Fig. 12)

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

In this fit we can identify the slope as the emission factor of a mixed fleet

EF_{f m}=1.4±0.1×1014veh−1km−1. Thereby this approach, the combination of flux mea-surements and traffic data, offers a method to estimate the emission factor averaged over the actual vehicle population for a prolonged sampling period. Our EF_{f m} com-pares reasonable well with previous studies, see Table 2, despite the large difference

5

in methods.

The correlation coefficient between w0N0 and traffic activity, R=0.79, which means that the explained variance of the flux is 62% (R2=0.62). The aerosol number source flux is hence largely caused by traffic related emissions. The offset F₀must represent all other sources, e.g. cooking, off-road vehicles, and other combustion sources. This

10

source flux F₀=14±18×106m−2 s−1 is not significant lager than zero considering the 95% confidence interval of the fit. The average emissions during daytime (between 06:00 and 20:00) in SW are ∼390×106m−2s−1, during this time F₀only contribute with 4% of the emissions.

To investigate the magnitude of the emissions from the construction area in the NE

15

sector, the measured fluxes can be compared with the fluxes calculated with the emis-sion factor for a fleet mix. The average particle flux during the whole measurement period increased with ∼430×106particles m−2 s−1, from F=99×106particles m−2 s−1 emitted from the traffic to a measured flux of 530×106particles m−2s−1.

The friction velocity U_∗ has the second best correlation R=0.449 with the aerosol

20

flux when the whole data set is analysed and the third best R=0.384 when the data set used for the parameterisation is analysed. When we looked at the diurnal average cycles, we also found further support for the relevance of U_∗. We would therefore like to combine the influence of U_∗with the linear dependency on TA, but keep the unit of the slope identical to that of EF_{f m}. We will therefore normalise U_∗with the average friction

25

velocity U_∗. Doing so, we have found

F = EF_{f m}T A U∗ U_∗

!0.4

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

Since a non-linear function of U_∗ with a power coefficient of 0.4 gives a bet-ter correlation than a linear dependency. This results in a somewhat smaller

EF_{f m}=1.3±0.1×1014veh−1 km−1, a larger value of F₀=26±16×106m−2 s−1, and an overall correlation R=0.81. The difference in EF_{f m} is just barely significant, while the difference in F₀ is not significant. Despite the combination of the two best correlated

5

factors, TA and U_∗, Eq. (7) is not able to explain a significantly larger part of the varia-tion than Eq. (6). Equavaria-tion (7) does a better job in predicting the daytime source flux, but overestimates the nighttime flux.

We are uncertain regarding the cause of the correlation between F and U_∗. A pos-sible interpretation of the effect of U_∗ is that it reflects variation in F due to more or

10

less efficient ventilation from the canopy, or turbulent transport up to the measurement level. In the later case, it would imply that the surface layer did not always extend to the measurement level. If Eq. (7) is averaged over the time,

 U∗

U_∗ 

will approach unity and Eqs. (7) and (6) will become identical, that is the variation in turbulent transport does not influence the average flux.

15

Another interpretation is related to the relationship between the actual footprints of the flux measurements, the real spatial distribution of sources and the sectors we use to derive TA. The size of the footprint will be inversely dependent on U_∗(e.g. Schuepp et al., 1990). When U_∗ changes, the contribution of different roads on the flux will change. This will certainly be responsible for part of the unexplained variability in the

20

data. However, if by accident there is a systematic trend in a large part of the data that cause the footprint to move in over stronger sources when U_∗ increases, it could contribute to the correlation. In either case, the inclusion of U_∗in Eq. (7) would tend to isolate the effect of the relationship between U_∗and F from TA. Equation (6) may cause an overestimate of EF_{f m} and the later smaller value resulting from Eq. (7) should be

25

the most appropriate.

Dorsey et al. (2002) included the mean wind speed in their parameterisation of the number concentration using a dependence on the mean horizontal wind speed of the

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

form ¯U0.3, which is proportional to U_∗0.3. This is close to our result. We have however found a weak preference for a power equal to 0.4. We have not been able to relate the emission factors directly to the stability, while Dorsey et al. (2002) related the number concentration to the stability with a complex function.

In Table 2 is a comparison with recently published emission factors based on different

5

methods (road tunnel studies, inverse modelling using street canyon measurements or road side measurements). As mentioned before they have the limitation not repre-senting the actual effective emission from a whole mixed vehicle fleet. Despite the differences in the methods, the agreement between our emission factor and literature data is fairly good.

10

As shown in Fig. 9 the fluxes are higher during weekdays then during holidays. One reason for this can be less traffic that would however not influence the emission fac-tor. Another reason can be a different mixture of vehicles with less HDV that could have an effect on the emission factor. For this reason the diurnally averages calculated for holidays and weekdays for the four 45◦ sectors, will be used in an attempt to find

15

out if different parameterisations can be found for weekdays and holidays. When the averaged fluxes are combined with the TA, using both weekdays and holidays, the cor-relation coefficient increases to R=0.85. One reason for this is that for a specific time of the day (and TA) now only one averaged emission is available, which has reduced the variation compared to earlier attempts, when the emissions were averaged over

20

one hour. The emission factor found using this method has the same size as with the method using one-hour means. When the weekdays and holidays are treated sepa-rately, the emission factor is significant lower during holidays then during weekdays. The emission factor decreased from EF_{f m}=1.4 1014veh−1 km−1 during weekdays to

EF_{f m}=0.4 1014veh−1 km−1 during holidays, i.e. only 30% of the weekday emission

25

factor. The probably reason for this is that the HDV traffic is less during holidays as shown before in Fig. 11, where the amount of HDV during holidays are only one third of the weekday traffic.

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

As a final attempt to find a parameterisation multiple linear regression techniques (Draper and Smith, 1981) will be used to examine the relationship between the LDV and HDV contribution to the emission factor. Now the urban aerosol number source flux can be written as

F = EF_LDVT A_LDV + EF_HDVT A_HDV + F₀ (8)

5

Two different emission factors are found one for LDV EF_LDV=0.3±0.3×1014 veh−1 km−1, which is not significant different from zero and one for HDV

EF_HDV=19.8±4.0×1014 veh−1km−1 and F₀=18±16×106 m−2 s−1, which is significant larger than zero considering the 95% confidence interval of the fit, R=0.83. This is in good agreement with earlier studies, se Table 2. During weekdays the HDV emit a

10

majority of the particles, between 70 to 80%, this differ over time and sector. During holidays when the HDV is reduced, around 50% of the particle emission came from HDV. In Table 3 the emissions factors are shown,

The regressions have been done for different stabilities, (not shown). No significant influences depending on the stratification are found. The higher correlation for only

15

neutral conditions (R=0.97 for the linear regression and R=0.99 for the multiple linear regression) does not correspond to better confidence since the numbers of data points have been reduced.

4. Summary and conclusions

The measured aerosol number flux shows a good quality (considering error

magni-20

tude and turbulent spectra) and the site is high enough above the sources not to be influenced by individual buildings, point sources or vehicles. A database including traffic activity has provided a unique possibility to take into account all traffic related sources influencing the emissions. After dividing the data into subsets according to wind direction, time of the day and traffic intensity it is concluded that the strongest

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

influence on the aerosol source fluxes is the traffic, the second strongest is the turbu-lent transport represented by the friction velocity U_∗. A linearity between the aerosol number flux and the traffic activity results in a first simple source parameterisation, Eq. (6), which explain 61% of the variation of the flux. The corresponding emission factor EF_{f m}=1.4±0.1×1014 compares reasonably well with previous values resulting

5

from other experiments. When we include the friction velocity U_∗in the parameterisa-tion, Eq. (7), in order to take into account the effects of various degrees of turbulent transport or canopy ventilation, the correlation remains the same. However, it results in a slightly smaller and just barely significant reduction of EF_{f m} to 1.3±0.1×1014. How-ever, the emission factor derived from Eq. (6) might be affected by periods when the

10

surface layer did not extend to the measurements height, while the emission factor de-rived from Eq. (7) to some degree is corrected for such effects. If a TA for a fleet mix is available the linear and simple source parameterisation in Eq. (6) is the one that should be applied in models, where of course the surface layer is perfectly defined, but one should probably apply it with the emission factor from Eq. (7). The correlation between

15

w0N0and U_∗may also be an effect of changes in the footprint with U_∗and corresponding spatial differences in TA.

Two separate methods show that during weekdays the HDV contributed to a ma-jority of the particle emissions. Different emission factors for weekdays and holidays are found. The ratio between these emission factors is comparable with the ratio

be-20

tween the HDV activity on holidays and weekdays. Finally multiple linear regression resulted in different emission factors for LDV EF_LDV=0.3±0.3×1014veh−1 km−1 and HDV EF_HDV=19.8±4.0×1014veh−1 km−1, Eq. (8). The fraction HDV of the total traffic activity came up to a level of ∼8% on the busiest roads, at this point up to ∼80% of the particle emissions, sometimes even more, originate from this type of traffic.

Con-25

struction within an urban area increases the average aerosol number emissions, due to more of road traffic and transports mostly performed with HDV. In this study, the emis-sion increased five times, compared to the assumed emisemis-sion from a fleet mix in the area influenced by construction. It is likely that even better correlation would have been

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

found if the hourly variation of the traffic activity of LDV and HDV had been measured in real time during the campaign.

Although the eddy covariance method has its own errors and difficulties, we have shown that the instrumental errors and methodological errors are relatively small, partly thanks to the high aerosol number concentrations in the urban environment. One of the

5

largest general problems with flux measurements is the question of how representative they are, since usually one measure only at one site, and this applies to our work as well as to most other studies. However, the problems associated with assumptions made when interpreting aerosol concentrations in terms of emissions, or when applying emission factors based on a very limited number of vehicles, are most probably larger.

10

Since the flux measurements in this work were made over an extended period and has a footprint on the order of a km2, the resulting emission factors represents a truly mixed vehicle population and various weather and traffic conditions. Only counting the largest roads and bridges, approximately 107vehicles passed through the area during the campaign. That would be impossible to achieve with the alternative methods.

15

In general, important improvements can be made if future separate data sets for other periods and locations, and hence somewhat different EF_{f m}, can be collected and analysed. A larger and better data base, would lead to a more reliable parame-terisation, and would help us to better understand the influence of meteorology, (here represented by U_∗), and to search for an eventual temperature influence on the source

20

fluxes. Size resolved fluxes and chemically specified fluxes (separating for instance soot from volatile compounds) would also be an important advancement in order to better understand the effects of this source, and to derive size resolved parameteriza-tions and source fluxes for different chemical compounds.

Acknowledgements. VINNOVA, “Verket f ¨or Innovationssystem”, Sweden, Project number

25

17418-1 and the Swedish Research Council G 5103-20005689/2000/369900 for financial sup-port. “Telia” for the possibility to use the tower. “City of Stockholm Environment and Health Administration” for access to the traffic database. L. B¨acklin and S. van Ekeren at MISU for help with equipment and help during the measurements.

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

References

Brokh ¨oj, E.: Dry deposition of aerosol particles in the urban environment, MSc thesis at De-partment of Meteorology, Stockholm University, 32 pp., 2003.

Buzorius, G., H ¨ameri, K., Pekkanen, J., and Kulmala, M.: Spatial variation of aerosol number concentration in Helsinki city, Atmos. Environ., 33, 553–565, 1999.

5

Buzorius, G.: Cut-off sizes and time constants of the CPC TSI 3010 operating at 1 to 3 lpm flow rates, Aerosol Sci. Technol., 35, 577–585, 2001.

Buzorius, G., Rannik, ¨U., M ¨akel ¨a, J. M., Keronen, P., Vesala, T., and Kulmala, M.: Vertical aerosol fluxes measured by the eddy covariance method and deposition of nucleation mode particles above a Scots pine forest in southern Finland, J. Geophys. Res., 105(D15), 19 905–

10

19 916, doi:10.1029/2000JD900108, 2000.

Buzorius, G., Rannik, ¨U., Nilsson, E. D., Vesala, T., and Kulmala, M.: Analysis of dry deposition velocity measurement techniques for particles smaller than 100 nm, J. Aerosol Sci., 34, 747– 764, 2003.

Dorsey, J. R., Nemitz, E., Gallagher, M. W., Fowler, D., Williams, P. I., Bower, K. N., and

15

Beswick, K. M.: Direct measurements and parameterisation of aerosol flux, concentration and emission velocity above a city, Atmos. Environ., 36, 791–800, 2002.

Draper, N. and Smith, H.: Applied Regression Analysis, Second Edition, Wiley, 1981.

Gidhagen, L., Johansson, C., Str ¨om, J., Kristensson, A., Swietlicki, E., Pirjolad, L., and Hans-son, H.-C.: Model simulation of ultrafine particles inside a road tunnel, Atmos. Environ., 37,

20

2023–2036, 2003.

Gidhagen, L., Johansson, C., Langner, J., and Olivares, G.: Simulation of NOx and ultrafine particles in a street canyon in Stockholm, Sweden, Atmos. Environ., 38, 2029–2044, 2004a. Gidhagen, L., Johansson, C., Omstedt, G., Langner, J., and Olivares, G.: Model simulations

of NOx and ultrafine particles close to a Swedish highway, 2004, Environ. Sci. Technol., 38,

25

6730–6740, 2004b.

Gidhagen, L., Johansson, C., Langner, J., and Foltescu, V.: Urban scale modelling of particle number concentration in Stockholm, Atmos. Environ., 39, 1711–1725, 2005.

Ibald-Mulli, A., Timonen, K. L., Peters, A., Heinrich, J., W ¨olke, G., Lanki, T., Buzorius, G., Kreyling, W. G., de Hartog, J., Hoek, G., ten Brink, H. M., and Pekkanen, J.: Effects of

30

Particulate Air Pollution on Blood Pressure and Heart Rate in Subjects with Cardiovascular Disease: A Multicenter Approach, EHP, 112, 3, 10.1289/ehp.6523, 2004.

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

IPCC: Climate Change 2001: The Scientific Basis, Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, edited by: Houghton, J. T., Ding, Y., Griggs, D. J., Noguer, M., van der Linden, P. J., Dai, X., Maskell, K., and Johnson, C. A., Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 881 pp., 2001.

5

Johnson, J. P., Kittelson, D. B., and Watts, W. F.: Source apportionment of diesel and spark ignition exhaust aerosol using on-road data from the Minneapolis metropolitan area, Atmos. Environ., 39, 2111–2121, 2005.

Ketzel, M., W ˚ahlin, P., Berkowicz, R., and Palmgren, F.: Particle and trace gas emission factors under urban driving conditions in Copenhagen based on street and roof level observations,

10

Atmos. Environ., 37, 2735–2749, 2003.

Kristensson, A., Johansson, C., Westerholm, R., Swietlicki, E., Gidhagen, L., Wideqvist, U., and Vesely, V.: Real-world traffic emission factors of gases and particles measured in a road tunnel in Stockholm, Sweden, Atmos. Environ., 38, 657–673, 2004.

Kirchstetter, T. W., Harley, R. A., Kreisberg, N. M., Stolzenburg, M. R., and Hering, S. V.:

On-15

road measurement of fine particle and nitrogen oxide emissions from light-and heavy-duty motor vehicles, Atmos. Environ., 33, 2955–2968, 1999.

Kittelson, D. B., Watts, W. F., and Johnson, J. P.: Nanoparticle emissions on Minnesota high-ways, Atmos. Environ., 38, 9–19, 2004.

Maricq, M. M., Podsiadlik, D. H., and Chase, R. E.: Gasoline vehicle particle size

distribu-20

tions: Comparison of steady state, FTP and US06 measurements, Environ. Sci. Technol., 33, 2007–2015, 1999.

Rannik, ¨U. and Vesala, T.: Autoregressive filtering versus linear detrending in estimation of fluxes by the eddy covariance method, Boundary-Layer Meteorol., 91, 259–289, 1999. Ruuskanen, J., Tuch, Th., Ten Brik, H., Peters, A., Khlystov, A., Mirme, A., Kos, G. P. A.,

25

Brunekreef, B., Wichmann, H. E., Buzorius, G., Vallius, M., Kreyling, W. G., and Pekkanen, J.: Concentrations of ultrafine, fine and PM2.5 particles in three European cities, Atmos. Environ., 35, 3729–3738, 2001.

Schack, C. J., Pratsinis, S. E., and Friedlander, S. K.: A general correlation for deposition of suspended particles from turbulent gases to completely rough surfaces, Atmos. Environ., 19,

30

953–960, 1985.

Schuepp, P. H., Leclerc, M. Y., Macpherson, J. I., and Desjardins, R. L.: Footprint prediction of scalar fluxes from analytical solutions of the diffusion equation, Boundary Layer Meteorol.,

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

50, 355–373, 1990.

Slinn, W. G. N.: Prediction for Particle Deposition to Vegetative Canopies, Atmos. Environ., 16, 1785–1794, 1982.

TSI: Model 3760A/3762 Condensation particle counter instruction manual, P/N 1933772, TSI Inc., St. Paul, MN, 2002.

5

WHO: Health aspects of air pollution with particulate matter, ozone and nitrogen dioxide, Re-port on a WHO working group, Copenhagen, WHO Regional Office for Europe, document EUR/03/5042688, 2003,http://www.euro.who.int/document/e79097.pdf.

WHO: Health aspects of air pollution – answers to follow-up questions from CAFE, Report on a WHO working group, Copenhagen, WHO Regional Office for Europe, document

10

EUR/04/5046026, 2004,http://www.euro.who.int/document/E82790.pdf.

W ˚ahlin, P., Palmgren, F., and Van Dingenen, R.: Experimental studies of ultrafine particles in streets and the relationship to traffic, Atmos. Environ., 35, 63–69, 2001.

Zhang, L., Gong, S. L., Padro, J., and Barrie, L.: A size-segregated particle dry deposition scheme for an atmospheric aerosol module, Atmos. Environ., 35, 549–560, 2001.

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

Table 1. (a) Statistics values on measurement results for all sectors, for temperature (T ), wind speed (U ), friction velocity (U_∗), aerosol particle number concentration (N), sensible heat flux w0_T0 _{and aerosol particle number flux w}0_N0_{. Included is also the aerosol particle number} concentration in a street canyon (Nstreet) and at a rooftop (Nroof).

Tower, all wind directions Street & rooftop

T U U_∗ N w0_T0 _w0_N0 _N street Nroof K ms−1 ms−1 cm−3 Wm−2 106m−2s−1 cm−3 cm−3 Mean 278 6.1 0.38 5599 27.7 260 70 845 10 070 Std 5 2.1 0.23 3004 65.4 422 46765 6120 Min 268.7 0.15 0.01 1296 −101.6 −351 8540 2480 Max 290.5 12.4 1.09 28 247 293.4 3317 345 100 68 075 Correlations: T 1.000 0.020 0.028 −0.192 0.274 −0.171 U 0.020 1.000 0.564 −0.208 −0.023 0.160 U∗ 0.028 0.564 1.000 0.031 0.363 0.449 N −0.192 −0.208 0.031 1.000 0.054 0.439 w0_T0 _{0.274 −0.023 0.363 0.054 1.000 0.267} w0_N0 _{−0.171 0.160 0.449 0.439 0.267 1.000} T U U_∗ N w0_T0 _w0_N0

ACPD

Full Screen / Esc

Print Version Interactive Discussion

EGU

Table 1. (b) Statistics values on measurement results for the SW and SE sectors used for the parameterisations, for temperature (T ), wind speed (U ), friction velocity (U_∗), aerosol particle number concentration (N), sensible heat flux w0_T0 _{and aerosol particle number flux w}0_N0_. In-cluded is also the aerosol particle number concentration in a street canyon (Nstreet) and at a

rooftop (Nroof).

Tower, wind direction 90◦–270◦ Street & rooftop

T U U_∗ N w0_T0 _w0_N0 _N street Nroof K ms−1 ms−1 cm−3 Wm−2 106m−2s−1 cm−3 cm−3 Mean 279 6.3 0.38 5218 30.2 153 70 845 10 070 Std 4 1.9 0.19 2252 59.1 216 46 765 6120 Min 269.9 1.2 0.02 1721 −59.0 1 8540 2480 Max 290.4 11.0 0.96 21 228 232.1 1185 345 100 68 075 Correlations: T 1.000 0.135 0.315 −0.155 0.396 0.024 U 0.135 1.000 0.463 −0.147 −0.098 0.030 U_∗ 0.315 0.463 1.000 −0.045 0.429 0.384 N −0.155 −0.147 −0.045 1.000 −0.067 0.448 w0_T0 _{0.396 −0.098 0.429 −0.067 1.000 0.274} w0_N0 _{0.024 0.030 0.384 0.448 0.274 1.000} T U U_∗ N w0_T0 _w0_N0